{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:QYEPAZX5NSBVVJIRPX54S4CODQ","short_pith_number":"pith:QYEPAZX5","schema_version":"1.0","canonical_sha256":"8608f066fd6c835aa5117dfbc9704e1c0385a6472a4a77d2d717f0cdf52ee6c2","source":{"kind":"arxiv","id":"2302.12022","version":3},"attestation_state":"computed","paper":{"title":"DoG is SGD's Best Friend: A Parameter-Free Dynamic Step Size Schedule","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Maor Ivgi, Oliver Hinder, Yair Carmon","submitted_at":"2023-02-08T18:28:48Z","abstract_excerpt":"We propose a tuning-free dynamic SGD step size formula, which we call Distance over Gradients (DoG). The DoG step sizes depend on simple empirical quantities (distance from the initial point and norms of gradients) and have no ``learning rate'' parameter. Theoretically, we show that a slight variation of the DoG formula enjoys strong parameter-free convergence guarantees for stochastic convex optimization assuming only \\emph{locally bounded} stochastic gradients. Empirically, we consider a broad range of vision and language transfer learning tasks, and show that DoG's performance is close to t"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2302.12022","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-02-08T18:28:48Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"8c18b303e8af9fa7145cec10e589b6df6e684fff0fafcac76b99da8fb016bd0d","abstract_canon_sha256":"29c0e7160494b68f6038941cca8a690713b0b9910bf4b746788a319f9e01914f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:31:09.476953Z","signature_b64":"rzgf/7Ivw43948pLueBFw8L5vUJ3+9TKc5v0yp6+3Syjzff7aXGNj/vgyFa3nKFY77JtQwXL4v8vi6kDpuwVDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8608f066fd6c835aa5117dfbc9704e1c0385a6472a4a77d2d717f0cdf52ee6c2","last_reissued_at":"2026-07-05T06:31:09.476449Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:31:09.476449Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DoG is SGD's Best Friend: A Parameter-Free Dynamic Step Size Schedule","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.LG","authors_text":"Maor Ivgi, Oliver Hinder, Yair Carmon","submitted_at":"2023-02-08T18:28:48Z","abstract_excerpt":"We propose a tuning-free dynamic SGD step size formula, which we call Distance over Gradients (DoG). The DoG step sizes depend on simple empirical quantities (distance from the initial point and norms of gradients) and have no ``learning rate'' parameter. Theoretically, we show that a slight variation of the DoG formula enjoys strong parameter-free convergence guarantees for stochastic convex optimization assuming only \\emph{locally bounded} stochastic gradients. Empirically, we consider a broad range of vision and language transfer learning tasks, and show that DoG's performance is close to t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.12022","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2302.12022/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2302.12022","created_at":"2026-07-05T06:31:09.476517+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.12022v3","created_at":"2026-07-05T06:31:09.476517+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.12022","created_at":"2026-07-05T06:31:09.476517+00:00"},{"alias_kind":"pith_short_12","alias_value":"QYEPAZX5NSBV","created_at":"2026-07-05T06:31:09.476517+00:00"},{"alias_kind":"pith_short_16","alias_value":"QYEPAZX5NSBVVJIR","created_at":"2026-07-05T06:31:09.476517+00:00"},{"alias_kind":"pith_short_8","alias_value":"QYEPAZX5","created_at":"2026-07-05T06:31:09.476517+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22669","citing_title":"Clipping the Price of Adaptivity at the Tail","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2412.14291","citing_title":"Projected gradient methods for nonconvex and stochastic smooth optimization: new complexities and auto-conditioned stepsizes","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24708","citing_title":"Scalable Hyperparameter-Divergent Ensemble Training with Automatic Learning Rate Exploration for Large Models","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ","json":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ.json","graph_json":"https://pith.science/api/pith-number/QYEPAZX5NSBVVJIRPX54S4CODQ/graph.json","events_json":"https://pith.science/api/pith-number/QYEPAZX5NSBVVJIRPX54S4CODQ/events.json","paper":"https://pith.science/paper/QYEPAZX5"},"agent_actions":{"view_html":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ","download_json":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ.json","view_paper":"https://pith.science/paper/QYEPAZX5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.12022&json=true","fetch_graph":"https://pith.science/api/pith-number/QYEPAZX5NSBVVJIRPX54S4CODQ/graph.json","fetch_events":"https://pith.science/api/pith-number/QYEPAZX5NSBVVJIRPX54S4CODQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ/action/storage_attestation","attest_author":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ/action/author_attestation","sign_citation":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ/action/citation_signature","submit_replication":"https://pith.science/pith/QYEPAZX5NSBVVJIRPX54S4CODQ/action/replication_record"}},"created_at":"2026-07-05T06:31:09.476517+00:00","updated_at":"2026-07-05T06:31:09.476517+00:00"}